AMD-GFX Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Jonathan Kim <jonathan.kim@amd.com>
To: <amd-gfx@lists.freedesktop.org>,
	<dri-devel@lists.freedesktop.org>, <Felix.Kuehling@amd.com>
Cc: Jinhuieric.Huang@amd.com
Subject: [PATCH 06/33] drm/amdgpu: add gfx9 hw debug mode enable and disable calls
Date: Thu, 25 May 2023 13:27:18 -0400	[thread overview]
Message-ID: <20230525172745.702700-6-jonathan.kim@amd.com> (raw)
In-Reply-To: <20230525172745.702700-1-jonathan.kim@amd.com>

Implement the per-device calls to enable or disable HW debug mode for
GFX9 prior to GFX9.4.1.

GFX9.4.1 and onward will require their own enable/disable sequence as
follow on patches.

When hardware debug mode setting is requested, waves will inherit
these settings in the Shader Processor Input's (SPI) Sequencer Global
Block (SQG). This means that the KGD must drain all waves from the SPI
into SQG (approximately 96 SPI clock cycles) prior to debug mode setting
to ensure that the order of operations that the debugger expects with
regards to debug mode setting transaction requests and wave inheritence
of that mode is upheld.

Also ensure that exception overrides are reset to their original state
prior to debug enable or disable.

Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
---
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c | 92 +++++++++++++++++++
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h |  9 ++
 2 files changed, 101 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
index 9fa9aab22fe9..64da5995170c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
@@ -649,6 +649,96 @@ int kgd_gfx_v9_wave_control_execute(struct amdgpu_device *adev,
 	return 0;
 }
 
+/*
+ * GFX9 helper for wave launch stall requirements on debug trap setting.
+ *
+ * vmid:
+ *   Target VMID to stall/unstall.
+ *
+ * stall:
+ *   0-unstall wave launch (enable), 1-stall wave launch (disable).
+ *   After wavefront launch has been stalled, allocated waves must drain from
+ *   SPI in order for debug trap settings to take effect on those waves.
+ *   This is roughly a ~96 clock cycle wait on SPI where a read on
+ *   SPI_GDBG_WAVE_CNTL translates to ~32 clock cycles.
+ *   KGD_GFX_V9_WAVE_LAUNCH_SPI_DRAIN_LATENCY indicates the number of reads required.
+ *
+ *   NOTE: We can afford to clear the entire STALL_VMID field on unstall
+ *   because GFX9.4.1 cannot support multi-process debugging due to trap
+ *   configuration and masking being limited to global scope.  Always assume
+ *   single process conditions.
+ */
+#define KGD_GFX_V9_WAVE_LAUNCH_SPI_DRAIN_LATENCY	3
+void kgd_gfx_v9_set_wave_launch_stall(struct amdgpu_device *adev,
+					uint32_t vmid,
+					bool stall)
+{
+	int i;
+	uint32_t data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+
+	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1))
+		data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_VMID,
+							stall ? 1 << vmid : 0);
+	else
+		data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_RA,
+							stall ? 1 : 0);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), data);
+
+	if (!stall)
+		return;
+
+	for (i = 0; i < KGD_GFX_V9_WAVE_LAUNCH_SPI_DRAIN_LATENCY; i++)
+		RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+}
+
+/*
+ * restore_dbg_registers is ignored here but is a general interface requirement
+ * for devices that support GFXOFF and where the RLC save/restore list
+ * does not support hw registers for debugging i.e. the driver has to manually
+ * initialize the debug mode registers after it has disabled GFX off during the
+ * debug session.
+ */
+uint32_t kgd_gfx_v9_enable_debug_trap(struct amdgpu_device *adev,
+				bool restore_dbg_registers,
+				uint32_t vmid)
+{
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+/*
+ * keep_trap_enabled is ignored here but is a general interface requirement
+ * for devices that support multi-process debugging where the performance
+ * overhead from trap temporary setup needs to be bypassed when the debug
+ * session has ended.
+ */
+uint32_t kgd_gfx_v9_disable_debug_trap(struct amdgpu_device *adev,
+					bool keep_trap_enabled,
+					uint32_t vmid)
+{
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
 void kgd_gfx_v9_set_vm_context_page_table_base(struct amdgpu_device *adev,
 			uint32_t vmid, uint64_t page_table_base)
 {
@@ -875,6 +965,8 @@ const struct kfd2kgd_calls gfx_v9_kfd2kgd = {
 	.get_atc_vmid_pasid_mapping_info =
 			kgd_gfx_v9_get_atc_vmid_pasid_mapping_info,
 	.set_vm_context_page_table_base = kgd_gfx_v9_set_vm_context_page_table_base,
+	.enable_debug_trap = kgd_gfx_v9_enable_debug_trap,
+	.disable_debug_trap = kgd_gfx_v9_disable_debug_trap,
 	.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
 	.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
index a241299f4fbc..9588ff055393 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
@@ -64,3 +64,12 @@ void kgd_gfx_v9_acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id,
 uint64_t kgd_gfx_v9_get_queue_mask(struct amdgpu_device *adev,
 				uint32_t pipe_id, uint32_t queue_id);
 void kgd_gfx_v9_release_queue(struct amdgpu_device *adev, uint32_t inst);
+void kgd_gfx_v9_set_wave_launch_stall(struct amdgpu_device *adev,
+					uint32_t vmid,
+					bool stall);
+uint32_t kgd_gfx_v9_enable_debug_trap(struct amdgpu_device *adev,
+				      bool restore_dbg_registers,
+				      uint32_t vmid);
+uint32_t kgd_gfx_v9_disable_debug_trap(struct amdgpu_device *adev,
+					bool keep_trap_enabled,
+					uint32_t vmid);
-- 
2.25.1


  parent reply	other threads:[~2023-05-25 17:28 UTC|newest]

Thread overview: 50+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-05-25 17:27 [PATCH 01/33] drm/amdkfd: add debug and runtime enable interface Jonathan Kim
2023-05-25 17:27 ` [PATCH 02/33] drm/amdkfd: display debug capabilities Jonathan Kim
2023-05-30 19:20   ` Felix Kuehling
2023-05-25 17:27 ` [PATCH 03/33] drm/amdkfd: prepare per-process debug enable and disable Jonathan Kim
2023-05-25 17:27 ` [PATCH 04/33] drm/amdgpu: add kgd hw debug mode setting interface Jonathan Kim
2023-05-25 17:27 ` [PATCH 05/33] drm/amdgpu: setup hw debug registers on driver initialization Jonathan Kim
2023-05-30 19:23   ` Felix Kuehling
2023-05-25 17:27 ` Jonathan Kim [this message]
2023-05-25 17:27 ` [PATCH 07/33] drm/amdgpu: add gfx9.4.1 hw debug mode enable and disable calls Jonathan Kim
2023-05-25 17:27 ` [PATCH 08/33] drm/amdkfd: fix kfd_suspend_all_processes Jonathan Kim
2023-05-25 17:27 ` [PATCH 09/33] drm/amdgpu: add gfx10 hw debug mode enable and disable calls Jonathan Kim
2023-05-25 17:27 ` [PATCH 10/33] drm/amdgpu: add gfx9.4.2 " Jonathan Kim
2023-05-25 17:27 ` [PATCH 11/33] drm/amdgpu: add gfx11 " Jonathan Kim
2023-05-25 17:27 ` [PATCH 12/33] drm/amdgpu: add configurable grace period for unmap queues Jonathan Kim
2023-05-30 19:28   ` Felix Kuehling
2023-05-25 17:27 ` [PATCH 13/33] drm/amdkfd: prepare map process for single process debug devices Jonathan Kim
2023-05-30 19:36   ` Felix Kuehling
2023-05-25 17:27 ` [PATCH 14/33] drm/amdgpu: prepare map process for multi-process " Jonathan Kim
2023-05-30 19:55   ` Felix Kuehling
2023-05-30 19:58     ` Kim, Jonathan
2023-05-25 17:27 ` [PATCH 15/33] drm/amdgpu: expose debug api for mes Jonathan Kim
2023-05-25 17:27 ` [PATCH 16/33] drm/amdkfd: add per process hw trap enable and disable functions Jonathan Kim
2023-05-30 20:04   ` Felix Kuehling
2023-05-25 17:27 ` [PATCH 17/33] drm/amdkfd: apply trap workaround for gfx11 Jonathan Kim
2023-05-25 17:27 ` [PATCH 18/33] drm/amdkfd: add raise exception event function Jonathan Kim
2023-05-30 20:07   ` Felix Kuehling
2023-05-25 17:27 ` [PATCH 19/33] drm/amdkfd: add send exception operation Jonathan Kim
2023-05-25 17:27 ` [PATCH 20/33] drm/amdkfd: add runtime enable operation Jonathan Kim
2023-05-30 20:11   ` Felix Kuehling
2023-05-25 17:27 ` [PATCH 21/33] drm/amdkfd: add debug trap enabled flag to tma Jonathan Kim
2023-05-25 17:27 ` [PATCH 22/33] drm/amdkfd: update process interrupt handling for debug events Jonathan Kim
2023-05-30 20:16   ` Felix Kuehling
2023-05-25 17:27 ` [PATCH 23/33] drm/amdkfd: add debug set exceptions enabled operation Jonathan Kim
2023-05-25 17:27 ` [PATCH 24/33] drm/amdkfd: add debug wave launch override operation Jonathan Kim
2023-05-30 20:21   ` Felix Kuehling
2023-05-25 17:27 ` [PATCH 25/33] drm/amdkfd: add debug wave launch mode operation Jonathan Kim
2023-05-30 20:22   ` Felix Kuehling
2023-05-25 17:27 ` [PATCH 26/33] drm/amdkfd: add debug suspend and resume process queues operation Jonathan Kim
2023-05-30 20:24   ` Felix Kuehling
2023-05-25 17:27 ` [PATCH 27/33] drm/amdkfd: add debug set and clear address watch points operation Jonathan Kim
2023-05-30 20:26   ` Felix Kuehling
2023-05-25 17:27 ` [PATCH 28/33] drm/amdkfd: add debug set flags operation Jonathan Kim
2023-05-30 20:30   ` Felix Kuehling
2023-05-25 17:27 ` [PATCH 29/33] drm/amdkfd: add debug query event operation Jonathan Kim
2023-05-25 17:27 ` [PATCH 30/33] drm/amdkfd: add debug query exception info operation Jonathan Kim
2023-05-25 17:27 ` [PATCH 31/33] drm/amdkfd: add debug queue snapshot operation Jonathan Kim
2023-05-25 17:27 ` [PATCH 32/33] drm/amdkfd: add debug device " Jonathan Kim
2023-05-30 20:31   ` Felix Kuehling
2023-05-25 17:27 ` [PATCH 33/33] drm/amdkfd: bump kfd ioctl minor version for debug api availability Jonathan Kim
2023-05-30 19:17 ` [PATCH 01/33] drm/amdkfd: add debug and runtime enable interface Felix Kuehling

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20230525172745.702700-6-jonathan.kim@amd.com \
    --to=jonathan.kim@amd.com \
    --cc=Felix.Kuehling@amd.com \
    --cc=Jinhuieric.Huang@amd.com \
    --cc=amd-gfx@lists.freedesktop.org \
    --cc=dri-devel@lists.freedesktop.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox